A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 6, 03:15 UTC
[CI] Split V1 Others into 3 separate jobs (#37016)
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Kevin H. Luu committed
Mar 23, 2026 at 15:44 UTC
2488a82f89b15ad2ebed12160dcc423d44210db2
1 file changed
+46
-10
.buildkite/test_areas/misc.yaml
+46
-10
@@ -2,11 +2,54 @@ group: Miscellaneous
2
depends_on:
3
- image-build
4
steps:
5
-- label: V1 Others
6
- timeout_in_minutes: 60
5
+- label: V1 Spec Decode
6
+ timeout_in_minutes: 30
7
source_file_dependencies:
8
- vllm/
9
- - tests/v1
9
+ - tests/v1/spec_decode
10
+ commands:
11
+ - export VLLM_WORKER_MULTIPROC_METHOD=spawn
12
+ # TODO: create another `optional` test group for slow tests
13
+ - pytest -v -s -m 'not slow_test' v1/spec_decode
14
+ mirror:
15
+ amd:
16
+ device: mi325_1
17
+ depends_on:
18
+ - image-build-amd
19
+
20
+- label: V1 Sample + Logits
21
+ timeout_in_minutes: 30
22
+ source_file_dependencies:
23
+ - vllm/
24
+ - tests/v1/sample
25
+ - tests/v1/logits_processors
26
+ - tests/v1/test_oracle.py
27
+ - tests/v1/test_request.py
28
+ - tests/v1/test_outputs.py
29
+ commands:
30
+ - export VLLM_WORKER_MULTIPROC_METHOD=spawn
31
+ - pytest -v -s v1/sample
32
+ - pytest -v -s v1/logits_processors
33
+ - pytest -v -s v1/test_oracle.py
34
+ - pytest -v -s v1/test_request.py
35
+ - pytest -v -s v1/test_outputs.py
36
+ mirror:
37
+ amd:
38
+ device: mi325_1
39
+ depends_on:
40
+ - image-build-amd
41
+
42
+- label: V1 Core + KV + Metrics
43
+ timeout_in_minutes: 30
44
+ source_file_dependencies:
45
+ - vllm/
46
+ - tests/v1/core
47
+ - tests/v1/executor
48
+ - tests/v1/kv_offload
49
+ - tests/v1/worker
50
+ - tests/v1/kv_connector/unit
51
+ - tests/v1/metrics
52
+ - tests/entrypoints/openai/correctness/test_lmeval.py
53
commands:
54
- uv pip install --system -r /vllm-workspace/requirements/kv_connectors.txt
55
- export VLLM_WORKER_MULTIPROC_METHOD=spawn
@@ -14,16 +57,9 @@ steps:
57
- pytest -v -s -m 'not cpu_test' v1/core
58
- pytest -v -s v1/executor
59
- pytest -v -s v1/kv_offload
17
- - pytest -v -s v1/sample
18
- - pytest -v -s v1/logits_processors
60
- pytest -v -s v1/worker
20
- # TODO: create another `optional` test group for slow tests
21
- - pytest -v -s -m 'not slow_test' v1/spec_decode
61
- pytest -v -s -m 'not cpu_test' v1/kv_connector/unit
62
- pytest -v -s -m 'not cpu_test' v1/metrics
24
- - pytest -v -s v1/test_oracle.py
25
- - pytest -v -s v1/test_request.py
26
- - pytest -v -s v1/test_outputs.py
63
# Integration test for streaming correctness (requires special branch).
64
- pip install -U git+https://github.com/robertgshaw2-redhat/lm-evaluation-harness.git@streaming-api
65
- pytest -v -s entrypoints/openai/correctness/test_lmeval.py::test_lm_eval_accuracy_v1_engine